Auto Guard

This repository provides solution that enable the software devleoper and tester for SDV to identify bugs/issues and deploy updated software in the Vehicle. the following Eclipse SDV projects:

• Ankaios - an embedded container and workload orchestrator targeted at automotive HPCs
• eCAL – a fast communication middleware following the pub-sub principle
• Ankaios Dashboard – a graphical user interface for the Ankaios orchestrator

The repository additionally uses the MQtt Broker service provided by:

• HiveMQ - MQTT Broker service built for flexibility, security, and scalability

Table of Contents

• Description
• Installation
• Roadmap

Links

• Ankaios docs
• Ankaios Dashboard
• Ankaios quickstart
• eCAL docs
• Podman
• What are devcontainers?

Description

This product enables a fleet manager or application developer to build and deploy multi-architecture containerized workloads via a CI/CD pipeline to either one vehicle or the entire fleet. The workload provided in this repository can detect predefined events and abnormal sensor values and report them to a cloud backend for a developer to evaluate and improve algorithms. the implemented example workload is set up to detect emergency breaking events and to create reports including attached vehicle data.

Architecture

The repository contains the following Ankaios workloads

• Log Publisher App
• Administrator workload

It also contains the developer tools

• Synthetic Data Generator
• CI/CD pipeline script and cloud the application
• Reporting API (backend) including a web GUI

Workload Administrator

The Workload Administrator connects to an MQTT broker hosted in the Cloud to recieve the deployment workload specification. It communicates with the Ankaios server over its control interface to perform the following actions on one of the HPCs:

• start containerized workloads
• update containerized workloads
• delete containerized workloads The Workload Administrator is started by Ankaios at the start of the vehicle.

Log Publisher App

This is an example workload which is available in this repository. The log publisher app subscribes to the topic eCAL measurment to retrieve dynamic vehicle sensor data to detect an emergency break event (trigger). It then communicates with the Reporting REST API to submit recorded sensor data in case of a trigger event detection.
The Log Publisher App is not started during the vehicle start but is created remotely.

Synthetic Data Generator

The Synthetic Data Generator is reponsible for artificially producing dynamic vehicle sensor data to create an emergency breaking scenario. It publishes its sensor data via eCAL to the topic eCAL measurement. The synthetic data can be used to test and validate the trigger event detection logic in the Log Publisher App. In the real test vehicle setup this component is replaced by in-vehicle components publishing real sensor data to this very topic.

Reporting API

The reporting API is our Cloud Application that will be used to both receive the reports (they'll be posted at POST /api/reports) and to visualize them in a user-friendly UI at /reports.

Installation

Running the application

Prerequisites

Docker and VScode with Dev Container extension. Open the Dev Container in VScode as usual. If it does not work for you, try the following steps:

docker build -t shift2sdv-dev:0.1 --target dev -f .devcontainer/Dockerfile .

Once done, you can start the container and run a bash shell in it:

docker run -it --privileged --name shift2sdv-dev -v <absolute/path/to>/challenge-shift-to-sdv:/workspaces/shift2sdv -p 25551:25551 --workdir /workspaces/shift2sdv shift2sdv-dev:0.1 /bin/bash

Reports dashboard (Outside the Dev Container)

From the project folder, navigate to:

cd backend/reports-backend

Run docker/podman compose to build and start the reports dashboard:

docker compose build
docker compose down
docker compose up

Leave the terminal open. The dashboard will be available at this link.

Ankaios cluster

Run the following command to build the container images and start the development deployment:

scripts/restart_shift2sdv

Building the applications with the CI/CD pipeline

Inside your Dev Container, log into your container registry with docker. In this repository, a dedicated dockerhub registry is used and already filled with the containerized workloads for demonstration purposes. If you want to update the applications, run

scripts/cicd-build-apps-multi-arch

to rebuild and reupload the multi architecture container images to the container registry.

Deploying the example application via MQTT

Replace REPORT_DASHBOARD_URI value inside scripts/cicd-last-step using the reports dashboard address. If you are running the reports dashboard locally, you can specify the host's ip address.

When done, you can deploy the Log Publisher App with

scripts/cicd-last-step

This sends out the MQTT message to start the workload on the target device. the script is made to stop the Log Publisher App after a minute again. You can also manually deploy (start or update) an application using MQTT by either sending an Ankaios Manifest to vehicle/{ID}/manifest/apply (ID can be anything or all for rollout to the entire fleet), or sending a JSON (example defined in scripts/cicd-last-step) to /vehicle/{ID}/workload/start. A workload can be stopped by sending the exact same manifest to vehicle/{ID}/manifest/delete or just sending the name of the workload (no JSON, just a string) to /vehicle/{ID}/workload/stop.

Use synthetic data

In order to test the Log Publisher App, we have created the synthetic_vdy app to publish synthetic measurements to the vehicle_dynamics_synthetic topic. Run

cd apps/synthetic_vdy
python3 -u ./synthetic_vdy_app.py

Critical reports will appear on the Reporting dashboard when the Log Publisher App is running.

Summary

• Workload Administrator application is ran using Ankaios at vehicle startup and listens for workload updates over MQTT
• A newly created containerized application Log Publisher App is built, pushed to a container registry and deployed to the vehicle over MQTT (which makes the Workload Administrator app and to start it over Ankaios)
• Via MQTT, the vehicle workload state change and regular updates of the vehicle are received
• On the Ankaios Dashboard, the running application is visible
• The Log Publisher App subscribes to eCAL data, detects an emergency brake and uploads the recording to the cloud over a REST API
• Emergency brake reports are visible in the GUI
• Via MQTT, the Log Publisher App's trigger condition can be updated, different reports show up on the GUI
• Via MQTT, the Log Publisher App can be stopped again
• The MQTT message received indicate it has in fact stopped
• On the Ankaios Dashboard, the running Log Publisher App application has disappeared

Roadmap

More features have been planned for possible future releases

• Dynamic Triggers Configuration: The current release of the "Log Publisher App" only supports recognition of "Emergency Break" events. Our plan to support more triggers is to move away from the current hard-coded approach and to instead provide the triggers configuration trough MQTT. References for this feature can be found at dynamic-triggers.py and sample-dynamic-triggers
• Attach Video to Report: The current release reports are formed mostly by vehicle_dynamics data. In order to provide even more detailed information for the manufacturers, we'd like to provide video data together with the report